Google has rolled out a set of Gemini updates aimed at schools, including textbook-verified answers, tighter classroom controls and features designed to give teachers more insight into how students are using AI. The company announced the changes ahead of the ISTE education technology conference.
The most concrete addition is a grounding feature that links Gemini’s responses to OpenStax textbook content. The intent is to reduce the confident, plausible, wrong explanations that have made teachers wary of allowing general-purpose chatbots into homework workflows.
Controls matter more than capabilities
For school deployments, the limiting factor is rarely what the model can do. It is what administrators can see and restrict. Google is adding settings that let educators configure how Gemini behaves in their classrooms and monitor usage patterns, responding to a year of feedback that AI tools arrived in schools before governance did.
This is the unglamorous work of making a consumer product fit an institutional environment. Districts buy on procurement and compliance grounds, not on benchmark scores.
The hallucination problem, narrowed
Grounding answers in a specific textbook does not eliminate errors, but it changes their character. A response that cites a known source can be checked against that source; a response synthesized from training data cannot. For subjects like introductory biology or algebra, where OpenStax coverage is strong, the trade-off favors reliability over breadth.
The limitation is equally clear. Many classroom questions fall outside any single textbook, and students will ask them anyway.
A market settling into norms
Google’s move follows a year in which schools have shifted from blanket bans toward supervised use. The companies that win education contracts will be the ones that treat teacher oversight as a core feature rather than an afterthought, and that accept slower rollouts in exchange for institutional trust.